not really an "issue" per se, but a question worth asking for the benefit of all: let's say that for many very good reasons, the proteomics lab team is running many separate batches of multiplexed samples, for a large experiment with many, many conditions (some/most of which end up being specific within each batch, but various consistent controls nicely repeated across batches). Additionally, the spectral post-processing (peptide matching, abundance estimations, etc) are also performed individuall per/within-batch, not "globally" across all batches -- such that some samples within each batch have the "usual" zeroes/dropouts observed for certain peptide fragments observed within that batch for other samples, but there are also peptide species that were only observed (in one-or-more samples) within a given batch, and consitutively unobserved (missing) in another given batch (and are thus not even present in the batch-specific abundance tables). It seems like this situation could be handled by proDA, e.g. if there were to be a second, independent set" of probabilistic dropout curves, shared by all samples sharing the saming batch, and separate/distinct from the sample-level dropout curves within each batch. Assume dozens of batches, not just one or two; and batch-correcting coefficients as part of the design matrix, orthogonal to the actual experimental design. Thoughts?
not really an "issue" per se, but a question worth asking for the benefit of all: let's say that for many very good reasons, the proteomics lab team is running many separate batches of multiplexed samples, for a large experiment with many, many conditions (some/most of which end up being specific within each batch, but various consistent controls nicely repeated across batches). Additionally, the spectral post-processing (peptide matching, abundance estimations, etc) are also performed individuall per/within-batch, not "globally" across all batches -- such that some samples within each batch have the "usual" zeroes/dropouts observed for certain peptide fragments observed within that batch for other samples, but there are also peptide species that were only observed (in one-or-more samples) within a given batch, and consitutively unobserved (missing) in another given batch (and are thus not even present in the batch-specific abundance tables). It seems like this situation could be handled by proDA, e.g. if there were to be a second, independent set" of probabilistic dropout curves, shared by all samples sharing the saming batch, and separate/distinct from the sample-level dropout curves within each batch. Assume dozens of batches, not just one or two; and batch-correcting coefficients as part of the design matrix, orthogonal to the actual experimental design. Thoughts?